Labeling

Summary

One of the data analysis tools provided by Kontomatik is based on machine learning models that analyze a vast numbers of transactions. Our proprietary algorithm can break down each transaction and assign appropriate labels from over 80 available categories (e.g., monthly, salary, compensation, loan, welfare, cash withdrawal, nightlife, rent).

Labels assignment rules

The label assignment is based on patterns identified by the model in historical data, which has been meticulously crafted by human engineers following strict rules for defining each label. Kontomatik analyzes various aspects of the data, including the transaction description (such as title, second party, and type), amount, and regularity. Each transaction is evaluated individually for all possible labels.

Labels are defined by specific criteria, including numerical values assigned to certain words, amounts, and rules for verifying dates. If a transaction is recognized by the dedicated machine learning classifier and meets all predefined algorithmic conditions, the label is assigned.

A single transaction can have multiple labels, also when none of the labels match with the transaction, no label will be assigned. The more detailed the account and transaction information provided, the more precise the labels become, and additional labels may be assigned.

Supported services

Labeling is supported for:

• Account Information Service
• PDF Statements Parsing
• PDF Confirmations Parsing

Depending on how and which Kontomatik product you use, labels will appear in the API responses or in the transaction list in Insight.

There is also the possibility to label your own data. This can be done through Owner Upload, which allows you to send your own owner and transactional data to Kontomatik servers. You can then download the aggregated data back with the labels already included.

Integration

If you have purchased the transaction labeling feature, labels will automatically appear next to the imported transactions:

• in case you are fetching the data via API: labels will appear within the <labels>* element in the API response within each transaction element,
• if you are viewing the data via Insight: labels will appear in the transactions list, next to each transaction,
• in JSON response of the PDF Confirmations endpoint: labels will appear within the analytics object for each transaction.

*If you are using the Vendors Recognition product, the structure will change. Please see our article regarding Vendors Recognition.

Business labeling

A separate labeling model for business accounts is available in Poland. This model has a distinct set of labels tailored to corporate finances.

The API response structure for business labeling is identical to the structure used for individual labeling.

If you would like to enable this service for your account, please contact our Data Science team at data@kontomatik.com.

List of labels

A full list of available labels can be found:

• In our documentation, with a description of each label.
• In the Kontomatik Coverage sheet, where you can find information regarding the availability of each label in each supported country.